Forestry & Forest Sciences · Postgraduate & Professional Practice · also Landscape Ecology, Ecosystem Modelling, Forest Landscape Management
Forests at landscape scale: patterns and processes, fragmentation and connectivity, disturbance and succession models, and decision-support for landscape planning. Builds on the Sustainability track's landscape-ecology-protected-areas with forest disturbance and forest landscape models.
In the exams: PG and PhD level; ICAR AICE-JRF 59 Unit 11; GATE EY Section 1 and 5
Forests are managed stand by stand, but many of the things people value in them, such as wide-ranging wildlife, clean water from catchments, resilience to fire and pests, and the capacity of species to move as the climate changes, depend on how stands are arranged across whole landscapes. Landscape ecology studies the spatial pattern of ecosystems over areas of many square kilometres and how that pattern affects, and is affected by, ecological processes. Ecosystem and landscape modelling turns this understanding into tools that project how forests, carbon, water and species distributions may change under different management, disturbance and climate scenarios.
This manual covers landscape structure (patches, corridors and the matrix), landscape metrics and how to interpret them, fragmentation and edge effects, connectivity and corridor design, disturbance and succession modelling at landscape scale, forest landscape models and scenario analysis, ecosystem process models for carbon, water and nutrient fluxes, species distribution modelling, and landscape planning with stakeholders. It builds on forest ecology, GIS and remote sensing, which are taught in their own subjects of this track, and on the general landscape ecology taught in the sustainability track; here the emphasis is on forests and on models.
This is postgraduate and doctoral material. Examinations for doctoral fellowships in forestry and graduate entrance and eligibility tests in ecology and environmental science ask about the patch–corridor–matrix model, island biogeography and species–area relationships, metapopulations, fragmentation and edge effects, disturbance regimes, landscape metrics, connectivity and species distribution models, with short numericals such as edge density, Shannon diversity of land cover, core area, fire rotation, steady-state proportions, decomposition constants and model evaluation statistics.
A landscape, in ecology, is a spatially heterogeneous area, usually several to many square kilometres, made up of interacting ecosystems. Landscape ecology asks how spatial pattern (the composition and arrangement of land-cover types) influences processes such as the movement of animals, seeds, water, nutrients and fire, and how those processes in turn create pattern. A central idea is that pattern and process are scale-dependent. Scale has two components: grain, the finest resolution of the data or of the organism's perception (a 30 m pixel, or the area over which a beetle responds to its surroundings), and extent, the total area or duration studied. Patterns that appear random at one scale may be clustered at another, and conclusions drawn at one scale cannot simply be transferred to another. Hierarchy theory views ecological systems as nested levels, each constrained by the level above and explained by processes at the level below: a stand within a watershed within a region. The same forest can be fine-grained for a moose and coarse-grained for a salamander, so landscape structure must always be defined relative to the organism or process of interest.
Forman and Godron described landscapes in terms of three elements. Patches are relatively homogeneous non-linear areas that differ from their surroundings, such as a woodlot in farmland, a burned area in forest, or a wetland. They arise in different ways: disturbance patches (a windthrow or fire), remnant patches (forest left after surrounding clearing), environmental resource patches (a wet hollow or rock outcrop that persists because of the site), and introduced patches (a plantation or settlement). Corridors are linear strips that differ from the land on both sides, such as riparian forest along a stream, a hedgerow, a road or a power line; they can conduct movement, act as barriers or filters, serve as habitat, or act as sources or sinks of organisms and materials. The matrix is the most extensive and most connected element, which exerts the strongest control over landscape function: in a heavily forested region the matrix is forest with patches of clearings, while in an agricultural region it is cropland with forest patches. The boundaries between elements, ecotones or edges, have their own conditions and species.
Every landscape can be described by this model, though in practice land cover often grades continuously, and alternative views, such as the gradient or continuum model, treat habitat as a continuous variable rather than as discrete patches. The choice affects which metrics and models make sense.
MacArthur and Wilson's theory of island biogeography, published in 1967, proposed that the number of species on an island reflects a balance between immigration, which falls with distance from the mainland, and extinction, which falls with island area. The theory was soon applied to habitat islands such as forest fragments in farmland. The species–area relationship, S = cAᶻ, where S is the number of species, A the area and c and z constants, summarises the increase of species with area; on a log–log plot it is a straight line with slope z. Values of z for isolated islands commonly fall between about 0.2 and 0.35, and for sample areas within continuous habitat they are often lower. With z = 0.25, a tenfold increase in area increases species by 10^0.25 ≈ 1.78 times, and doubling the area adds about 19%. The relationship underlies rough estimates of species loss after habitat loss, though such estimates are uncertain because the matrix is not an ocean: many species use or cross it, and extinction may lag behind habitat loss by decades.
Landscape metrics quantify pattern from a categorical map of land-cover classes. Composition metrics describe what is present and how much, without regard to location: the proportion of the landscape in each class, the number of classes (richness) and diversity indices. The Shannon diversity index H′ = −Σ pᵢ ln pᵢ, where pᵢ is the proportion of class i, increases with the number of classes and the evenness of their shares: for three classes covering 50%, 30% and 20%, H′ = −(0.5 ln 0.5 + 0.3 ln 0.3 + 0.2 ln 0.2) ≈ 1.03, and evenness, H′ ÷ ln m with m = 3 classes, is 1.03 ÷ 1.10 ≈ 0.94. Configuration metrics describe the spatial arrangement: patch size and number, shape, edge, isolation, aggregation and connectivity. Metrics are calculated at three levels: for each patch, for each class (all patches of forest, for example), and for the whole landscape.
Patch area and its distribution (mean patch size, largest patch index, the share of the landscape in the largest patch) describe the grain of the mosaic. Edge density is total edge length divided by landscape area, usually in metres per hectare: 24 km of forest edge in a 1,200 ha landscape gives 24,000 ÷ 1,200 = 20 m per ha. Shape indices compare a patch's perimeter with that of the most compact shape of the same area: in vector form, shape index = P ÷ (2√(πA)), which equals 1 for a circle and increases as shapes become more convoluted; a 10 ha patch (100,000 m²) with a 1,500 m perimeter has a shape index of about 1.34. In raster form the standard is a square, and the index is P ÷ (4√A). The perimeter–area fractal dimension describes shape complexity across patches of different sizes. Core area is the area remaining after removing an edge zone of specified depth, and is often more meaningful for interior species than total area. Nearest-neighbour distance between patches of the same class measures isolation. Contagion and aggregation indices measure the tendency of classes to clump. Effective mesh size, m_eff = (1 ÷ A_total) × Σ Aᵢ², expresses the probability that two random points lie in the same patch, scaled to area; for patches of 40, 30 and 10 km² in a 100 km² landscape, m_eff = (1,600 + 900 + 100) ÷ 100 = 26 km². It is used to track fragmentation by roads.
Metrics are easy to compute and easy to misuse. Many are strongly correlated with one another and with the proportion of the focal class, so a large set of metrics often contains only a few independent dimensions; principal component analysis or careful selection of a small set is advised. Values depend on the grain and extent of the map, on the classification scheme (how many classes and how they are defined), on whether patches are defined by the four-neighbour or eight-neighbour rule in raster data, and on the minimum mapping unit. The same landscape mapped at 30 m and at 1 m can give very different patch numbers and edge densities. Metrics should therefore be compared only between maps made in the same way, and should be chosen because they relate to a process of interest, such as dispersal distance for a species or edge length for an edge-sensitive bird, rather than because software produces them. Metrics describe pattern; showing that pattern affects a process requires field data or models.
A neutral landscape model is a map generated by a known random process, used as a null model against which real landscapes and ecological responses can be compared. Gardner and colleagues introduced simple random maps in 1987, in which each cell is habitat with probability p independently of its neighbours; later work added fractal and hierarchical neutral maps in which habitat is spatially clumped to a controlled degree. Such maps show how metrics behave purely as a function of habitat amount and clumping: on random maps, the number of patches peaks at intermediate p, the largest patch grows abruptly near the percolation threshold, and edge density is greatest near p = 0.5. Comparing a real landscape with neutral maps of the same habitat proportion tells the analyst whether its pattern is more fragmented or more aggregated than chance, and simulation of dispersal or fire on neutral maps isolates the effect of configuration from that of habitat amount. Neutral models therefore help to avoid the common error of attributing to configuration a result that is simply a consequence of how much habitat is present.
Fragmentation is the process by which a large, continuous area of habitat is broken into smaller, more isolated pieces. It usually proceeds through recognisable stages: perforation (small openings in continuous forest), dissection (division by roads or other linear features), fragmentation proper (breaking into separate patches), shrinkage of the patches and attrition, the loss of some patches altogether. Fragmentation in the broad sense combines two different things: habitat loss, the reduction in total habitat area, and fragmentation per se, the breaking apart of habitat for a given amount. Reviews by Fahrig and others have argued that habitat loss has much larger and more consistently negative effects on biodiversity than fragmentation per se, whose effects can be negative, neutral or even positive depending on species and context; this remains debated, but it has made clear that the two must be distinguished in study design.
Consequences of fragmentation include smaller populations, which are more vulnerable to demographic and genetic problems; isolation, which reduces recolonisation and gene flow; loss of area-sensitive and interior species; increased exposure to edge effects; and changed disturbance regimes, such as fire entering from surrounding farmland. Responses may be delayed: an extinction debt exists when species persist for a time in fragments too small or isolated to support them in the long run. Quantitative models of percolation show that random landscapes lose connectivity abruptly when habitat falls below a threshold; for a random map on a square grid with four-neighbour connections, a spanning cluster appears when about 59% of cells are habitat, though real landscapes, which are not random, and organisms that cross gaps, show thresholds at other levels.
An edge is the boundary between two habitats, and edge effects are the changes in physical conditions and biological processes near it. At a forest edge adjoining open land, more light and wind penetrate, air and soil are warmer and drier, and vapour pressure deficit is higher; these abiotic changes lead to increased tree mortality and damage, changes in plant species composition (more light-demanding and weedy species, lianas in tropical forests), and changes in animal communities. Nest predation and brood parasitism of forest birds are often higher near edges, and edges can be entry points for invasive species, livestock, hunters and fire. Long-term experimental studies of forest fragments, such as those established in the central Amazon in 1979, have shown that some edge effects penetrate tens of metres and others a few hundred metres into the forest, and that small fragments become dominated by edge conditions.
The importance of edge effects for a patch depends on its size and shape. If an edge effect extends 100 m into a square patch of 100 ha (1,000 m by 1,000 m), the core area is 800 m by 800 m = 64 ha, so 36% of the patch is edge; a 10 ha square patch with the same edge depth (about 316 m on a side) keeps only about 1.35 ha of core, and a long narrow patch may have none. Edges are not always bad: some species depend on them, and edge habitat was historically managed to favour game. The management goal is usually to maintain enough core habitat for interior species, by keeping large compact patches, softening edges with transitional vegetation and locating clearings and roads to minimise edge creation.
Connectivity is the degree to which a landscape facilitates or impedes movement among resource patches. Structural connectivity describes the physical arrangement of habitat, such as whether forest patches touch or are linked by hedgerows, without reference to any organism. Functional connectivity describes how a particular species actually moves, and depends on its dispersal ability, behaviour and the permeability of the matrix: a gap of 200 m of pasture may be no obstacle to a bird but a complete barrier to a forest salamander. Connectivity matters for daily and seasonal movements, dispersal of young, gene flow, recolonisation of patches after local extinction and, increasingly, the shifting of species ranges as the climate changes.
Metapopulation theory describes a set of local populations in habitat patches linked by dispersal, each of which may go extinct and be recolonised. In Levins' simple model, the fraction of occupied patches changes as dp/dt = c p (1 − p) − e p, where c is the colonisation rate and e the extinction rate; at equilibrium p* = 1 − e ÷ c, so with e = 0.2 and c = 0.5 about 60% of patches are occupied, and the metapopulation persists only if c exceeds e. Reducing connectivity lowers c and can drive a metapopulation to extinction even when every patch is still present. Landscape genetics uses genetic data to measure functional connectivity, testing whether genetic distance among populations is better explained by geographic distance (isolation by distance) or by landscape resistance.
Graph theory represents habitat patches as nodes and potential movement between them as links, weighted by distance or probability of dispersal. Indices such as the probability of connectivity and the integral index of connectivity measure how well the network as a whole is connected and rank each patch or link by how much its loss would reduce connectivity, which identifies critical stepping stones. Resistance surfaces assign each cell a cost of movement, based on land cover, roads, slope or expert opinion, and least-cost path analysis finds the cheapest route between patches; corridor maps show the zone within a set cost of that path. Circuit theory, adapted from electrical circuits, treats the landscape as a conductive surface and models movement as current flowing between patches; it identifies all routes, not just the cheapest, and highlights pinch points where movement is concentrated. Resistance values are the weakest part of most analyses and should be based on movement or genetic data where possible and tested by sensitivity analysis.
Corridors are protected or restored linear habitats that link patches. Their effectiveness depends on width (enough to provide interior conditions or at least shelter for the target species, which for many forest species means hundreds of metres rather than a few), continuity (gaps and road crossings are often the limiting points), habitat quality, length and the species' behaviour. Riparian corridors along streams are often the backbone of forest connectivity networks because they also protect water quality. Stepping stones, unconnected patches close enough to be crossed, can serve species that move through the matrix. Road crossings for wildlife, such as underpasses and overpasses with fencing to guide animals, reduce mortality and restore movement. Corridors also have potential costs: they may spread fire, disease, pests or invasive species, they can be edge-dominated traps if too narrow, and the land they occupy has an opportunity cost. Evidence from experiments and observational studies shows that corridors generally increase movement between patches, though the size of the effect varies with species and design, so corridor proposals should state their target species and expected movement.
A disturbance is a relatively discrete event that disrupts ecosystem structure and changes resources or the physical environment, such as fire, windstorm, insect outbreak, flood, landslide or harvesting. A disturbance regime describes the pattern of disturbances in a landscape over time: type, frequency or return interval, size, severity (the amount of vegetation killed or biomass removed), intensity (the energy released, for fire), seasonality and spatial pattern, and the interactions among disturbances. Fire rotation, or fire cycle, is the time needed to burn an area equal to the study area: if 1,000 ha burn per year on average in a 50,000 ha landscape, the fire rotation is 50 years, although some places burn several times in that period and others not at all. Under a constant, age-independent probability of burning, the stand-age distribution follows a negative exponential: the proportion of the landscape older than age t is e^(−t ÷ b), where b is the mean fire interval; with b = 100 years, about e^(−2) ≈ 13.5% of the landscape is older than 200 years. Real regimes depart from this because fuel builds with age, climate varies, and people suppress or ignite fires.
Disturbance regimes have varied over centuries, and the historical range of variability, the range of conditions under past regimes, is used as a reference for management, with the caveat that future climate may not permit a return to past conditions. The intermediate disturbance hypothesis proposes that species diversity peaks at intermediate frequencies or intensities of disturbance, though empirical support is mixed. In large landscapes with frequent small disturbances, the proportion in each successional stage may stay roughly constant even though individual patches change, a shifting-mosaic steady state; where disturbances are large relative to the landscape, as with large crown fires, no such equilibrium exists.
Succession models project how vegetation changes after disturbance. Markov models divide vegetation into states, such as young, mature and old forest, and use a matrix of transition probabilities per time step; repeated multiplication of the state vector by the matrix projects composition, and the system approaches a steady state that depends only on the transition probabilities. In a two-state model where young forest becomes old with probability 0.1 per year and old forest is reset to young by disturbance with probability 0.02 per year, the steady-state share of old forest is 0.1 ÷ (0.1 + 0.02) ≈ 83%. Markov models are simple and transparent but assume that transitions depend only on the current state.
Gap models simulate the establishment, growth and death of individual trees on small plots about the size of a canopy gap, following the JABOWA model of Botkin and colleagues (1972) and its many descendants; they represent competition for light, temperature and moisture limits and have been widely used to project responses to climate change. Individual-based spatially explicit models simulate the positions of trees and light in three dimensions. Cellular automata represent the landscape as a grid in which each cell's state changes according to rules involving its neighbours, a structure used for fire spread, insect outbreaks and land-use change. Choosing among these depends on the question, the data available and the scale.
Forest landscape models (FLMs) are spatially explicit simulation models that project forest change over large areas, typically thousands to millions of hectares, and long periods, decades to centuries, by combining stand-level processes with landscape processes. Their core components are succession and growth within each cell (often represented by species and age cohorts or by biomass), seed dispersal between cells, natural disturbances (fire, wind, insects, browsing) that spread across the landscape, and management (harvesting, planting, fuel treatments). Some FLMs, such as the LANDIS family developed in North America and now used on several continents, simulate species presence and age cohorts with dispersal and disturbance modules; others link process-based growth models in each cell to landscape disturbance. Cell sizes typically range from tens of metres to a kilometre and time steps from one to ten years. FLMs can be coupled with climate projections by making establishment, growth and disturbance depend on climate variables.
FLMs are used to explore questions that stand-level models cannot answer: how a change in harvest pattern affects habitat connectivity and fire spread; how climate change might shift species composition across a region; what mix of fuel treatments best reduces area burned; or how insect outbreaks and salvage logging interact. Their strength lies in integrating many processes; their weakness is the number of parameters, many uncertain, and the difficulty of validating projections for decades ahead.
A scenario is a plausible, internally consistent description of how the future might unfold, not a prediction. Scenario analysis compares outcomes under several alternatives, for example a business-as-usual management scenario, a scenario with larger reserves and longer rotations, and a scenario of intensive plantation forestry, each perhaps under more than one climate projection. Comparing outcomes such as timber supply, carbon stocks, old-forest area, habitat for focal species and fire risk across scenarios shows trade-offs and robust strategies, those that perform acceptably across many possible futures. Exploratory scenarios ask what could happen; normative scenarios work back from desired futures. Storylines developed with stakeholders help define scenarios that matter to decision makers.
Model uncertainty has several sources: parameter uncertainty (values not known precisely), structural uncertainty (the model's equations may be wrong or incomplete), input and scenario uncertainty (future climate and management), and stochastic variation (random fire and dispersal events). Good practice runs many replicate simulations of stochastic models and reports their range, performs sensitivity analysis to identify which parameters most influence outputs, compares models where possible (ensembles), and is transparent about assumptions. Verification checks that the computer code does what the model description says; validation tests model outputs against independent observations, such as historical fire records, inventory data or reconstructed landscape history. Pattern-oriented modelling tests whether a model reproduces several observed patterns at different scales at the same time, which constrains structure and parameters more strongly than a single fit. Documentation protocols such as ODD (overview, design concepts and details) make individual-based and landscape models reproducible.
Ecosystem process models simulate the flows of carbon, water and nutrients through forests from physiological and physical principles. Gross primary production (GPP) is the total carbon fixed by photosynthesis. Autotrophic respiration (Ra) by plants returns part of it, and the remainder is net primary production, NPP = GPP − Ra, the carbon available for growth of wood, leaves, roots and reproduction. Heterotrophic respiration (Rh) by decomposers and animals returns carbon from litter and soil, and net ecosystem production is NEP = NPP − Rh. Including losses from disturbance, harvest and leaching gives net biome production (NBP), the long-term carbon balance relevant to climate. For many forests NPP is roughly half of GPP, though the ratio varies with age, climate and species and is debated.
Light-use efficiency models estimate production from absorbed light: GPP = ε × fAPAR × PAR, where PAR is incoming photosynthetically active radiation, fAPAR the fraction absorbed by the canopy (estimated from leaf area index or satellite vegetation indices), and ε the light-use efficiency. With PAR of 2,500 MJ m⁻² per year, fAPAR of 0.8 and ε of 1.0 g C per MJ, GPP is 2,000 g C m⁻² per year, or 20 t C per ha per year. The 3-PG model applies this approach to plantation forestry, with modifiers for temperature, vapour pressure deficit, soil water, nutrition and age. More detailed biogeochemical models simulate photosynthesis with leaf biochemistry, allocate carbon among plant parts and track soil organic matter in pools that turn over at different rates. Eddy covariance flux towers, organised in regional and global networks, measure net ecosystem exchange of CO₂, water and energy continuously and are the main data for testing such models.
The water balance of a forested catchment over a period is P = ET + Q + ΔS, where P is precipitation, ET evapotranspiration (transpiration, interception loss and soil evaporation), Q runoff and ΔS the change in storage in soil and groundwater. Process models compute ET from radiation, temperature, humidity, wind and canopy conductance, typically with the Penman–Monteith equation, and route water through canopy, soil layers and streams. Water-use efficiency, carbon gained per unit water transpired, links the carbon and water models.
Nutrient cycling models track nitrogen and other elements through uptake, litterfall, decomposition, mineralisation, immobilisation, leaching and gaseous losses. Litter decomposition is often described by Olson's exponential model, Xₜ = X₀e^(−kt), where k is the decomposition constant; with k = 0.5 per year, e^(−1) ≈ 37% of the litter remains after two years, and the half-life is ln 2 ÷ k ≈ 1.4 years. Values of k range from well above 1 per year in warm humid tropical forests to well below 0.1 per year in cold boreal forests, which is why boreal soils accumulate organic matter. Models of soil organic matter divide it into pools with fast, slow and passive turnover.
Process-based models are preferred over empirical growth and yield models when conditions move outside the range of past data, such as under elevated CO₂, novel climates or new silvicultural regimes, because they represent mechanisms rather than historical relationships. They need more inputs and parameters and are harder to calibrate, and their projections are only as good as their representation of processes such as acclimation, drought mortality and nutrient limitation, which remain uncertain. Hybrid models combine empirical growth relationships with process-based modifiers. Dynamic global vegetation models extend ecosystem models to the globe, simulating plant functional types, and are part of Earth system models used for climate projections. For forest management, process models are used to compare species and silvicultural options under future climates, to estimate carbon sequestration and water yield, and to understand productivity differences among sites.
Species distribution models (SDMs), also called ecological niche or habitat suitability models, relate known occurrences of a species to environmental variables in order to map where conditions are suitable. They rest on niche theory: Hutchinson defined the fundamental niche as the set of environmental conditions in which a species can persist, and the realised niche as the narrower set it actually occupies, given competition, predation and dispersal limits. Correlative SDMs estimate something close to the realised niche from observed distributions. Methods include climatic envelope models, which define the range of each variable where the species occurs; regression methods such as generalised linear and generalised additive models fitted to presence–absence data; machine-learning methods such as random forests and boosted regression trees; and maximum entropy modelling, widely used for presence-only data, which compares occurrences with background points sampled from the study area. Predictors typically include climate variables, soil, topography and land cover, chosen for their biological relevance and screened for strong correlation with one another.
Model performance is assessed by comparing predictions with occurrence data not used to fit the model. For presence–absence predictions, a confusion matrix counts true positives, false negatives, true negatives and false positives. Sensitivity is the proportion of presences correctly predicted, specificity the proportion of absences correctly predicted, and the true skill statistic TSS = sensitivity + specificity − 1 ranges from −1 to +1, with 0 meaning no better than random. With 40 true positives, 10 false negatives, 35 true negatives and 15 false positives, sensitivity is 0.80, specificity 0.70 and TSS 0.50. The area under the receiver operating characteristic curve (AUC) summarises performance across all thresholds; 0.5 is random and 1 perfect. AUC is affected by the extent of the study area and the prevalence of the species, so a high AUC for a species with a small range in a large study area can be misleadingly reassuring. Because occurrences close together share environmental conditions, random splits of data into training and test sets overestimate performance; spatial block cross-validation gives more honest estimates. Occurrence records from herbaria, museums and citizen science are often biased towards roads, cities and protected areas, and such sampling bias must be corrected, for example by choosing background points with the same bias.
SDMs are used to map habitat for conservation planning, to identify areas at risk of invasion by alien species, to guide surveys for rare species, to choose sites for restoration and to project species ranges under future climates. Projection to new times or places assumes that the species is in equilibrium with climate, that the relationships hold outside the fitted range of conditions, and that dispersal can keep up; all three may fail. Novel climates, combinations of conditions with no present analogue, are a particular problem, and maps of extrapolation should accompany projections. Mechanistic SDMs model physiological tolerances directly and can be combined with correlative ones. In forestry, SDMs and related climate-matching methods inform choices of species and seed sources for planting under climate change, including assisted migration, and assessments of which forest types may be most at risk. Ensembles of several methods and climate models express uncertainty, and results should be presented as ranges of possibilities, not single maps.
Landscape planning seeks to balance multiple objectives, such as timber, biodiversity, water, carbon, livelihoods and recreation, across an area large enough to contain their trade-offs. The landscape approach, summarised in ten principles by Sayer and colleagues in 2013, emphasises continual learning and adaptive management, common concern entry points, multiple scales, multifunctionality, multiple stakeholders, negotiated and transparent change logic, clarification of rights and responsibilities, participatory and user-friendly monitoring, resilience, and strengthened stakeholder capacity. A long-running debate contrasts land sharing, in which production and conservation are integrated on the same land, with land sparing, in which high-yield production on some land allows other land to be protected; evidence suggests that the best balance depends on the species, the yield–density relationships and the governance context, and most real landscapes combine both.
Forest landscape restoration aims to regain ecological integrity and enhance human well-being across deforested or degraded landscapes, through a mosaic of interventions such as natural regeneration, enrichment planting, agroforestry and plantations. The Bonn Challenge, launched in 2011, set a global goal of bringing 150 million ha into restoration by 2020, extended by the New York Declaration on Forests in 2014 to 350 million ha by 2030.
Systematic conservation planning, set out by Margules and Pressey in 2000, chooses areas for protection by explicit, quantitative and transparent steps: compile data on biodiversity and costs; set conservation targets (for example, a share of each forest type or a number of populations of each species); review existing reserves for how far they meet targets; select additional areas efficiently; implement; and monitor. Key principles are representativeness (all types of biodiversity are included), complementarity (each new area adds features not already protected), efficiency (targets are met at least cost or area), irreplaceability (some sites are essential because the features they hold occur nowhere else), adequacy (enough area to ensure persistence) and comprehensiveness, often summarised as comprehensive, adequate and representative (CAR) reserve systems. Reserve-selection algorithms, often using simulated annealing, find near-optimal sets of sites that meet all targets while minimising cost and, optionally, boundary length to favour compact reserves.
Landscape decisions involve trade-offs among objectives and among people. Spatial decision support combines maps of ecosystem services and biodiversity, models of how they respond to land-use options, and tools for comparing scenarios. A trade-off or production-possibility frontier plots the best attainable combinations of two objectives, such as timber and carbon or timber and habitat; options below the frontier are inefficient, because another option gives more of one objective without less of the other, while choosing among points on the frontier is a value judgement for stakeholders. Participatory modelling and scenario workshops bring stakeholders into defining objectives, building models and interpreting results, which improves both the knowledge base and the legitimacy of decisions; companion modelling uses role-playing games and agent-based models with communities. Adaptive management treats plans as hypotheses, monitors outcomes and adjusts, which is essential where models are uncertain. Good practice records who was involved, what assumptions and values were used, and how disagreements were resolved, so that decisions can be revisited as conditions change.
Plans are only as good as the monitoring that tests them. Landscape monitoring combines repeated land-cover mapping from satellite imagery, which tracks the area, configuration and connectivity of forest types, with field indicators such as occupancy of focal species, regeneration, water quality and the condition of corridors and restoration sites. Indicators should be chosen in advance, linked to the plan's objectives, measured consistently, and analysed with methods that separate real change from mapping error; area change in particular should be estimated from reference samples rather than by comparing two maps of different quality. Participatory monitoring, in which communities record forest use, fires, wildlife or restoration survival, adds local detail and builds ownership, provided methods are simple and data are checked. Regular reporting back to stakeholders, with honest statements of what is and is not working, is what turns a landscape plan into adaptive management.
A postgraduate student uses this subject to compute and interpret landscape metrics from classified maps, to build species distribution models with honest evaluation, to analyse connectivity for a focal species with least-cost or circuit methods, and to run simple disturbance or succession models. In examinations the student must explain the patch–corridor–matrix model, island biogeography, metapopulations, edge effects and disturbance regimes, and calculate edge density, Shannon diversity, core area, fire rotation, steady-state proportions, decay constants and TSS. The commonest errors are comparing metrics from maps of different resolution and evaluating SDMs on random rather than spatial splits.
A forest planner or ecologist five years into practice designs harvest patterns that limit fragmentation and maintain core habitat, maps riparian and upland corridors, checks proposed roads against connectivity models, uses species distribution maps to target surveys for rare species, and runs landscape scenarios to show how alternative plans affect habitat, carbon and fire risk. Daily judgements include which edge depth to assume for a target species, whether a corridor is wide enough to be useful, and how much confidence to place in a model projection when presenting it to managers or the public.
A senior professional with twenty-five years of experience leads landscape-level planning for a region, a company estate or a protected-area network, negotiating trade-offs among timber, conservation, water, carbon and community needs. They decide which models and scenarios are fit for a decision, commission landscape analyses, approve reserve designs and restoration priorities, defend plans in public consultations and in court, and set up monitoring so that plans can be adapted. They must recognise when model results are robust across scenarios and when they hinge on uncertain assumptions.
A researcher with a PhD and long experience develops forest landscape and ecosystem models, tests them against long-term data, designs connectivity and conservation-planning methods, and studies how disturbance regimes, climate change and management interact across landscapes. Such researchers lead model intercomparisons and ensembles, advise national and international processes on restoration, connectivity and climate adaptation, review the use of models in policy, supervise doctoral students and examine fellowship candidates. They also decide which long-term datasets and model platforms deserve continued investment, and they train practitioners to interpret model outputs critically rather than as forecasts.
This manual explains concepts and simple forms of the models; it does not teach any software or the full equations of particular forest landscape or biogeochemical models. Values such as edge-effect depths, species–area exponents, NPP:GPP ratios and decomposition constants vary widely with species, region and method and are given only as illustrations. The debate on fragmentation per se and on land sharing versus sparing is unresolved and is presented as such. Agent-based models of human land use and remote-sensing methods for landscape mapping are only touched on.
Original practice questions written to the syllabus and pattern of the exams named in each question's examTags. They are not past papers and do not reproduce any board's questions.
These notes are free to read here, with no account.
What the Laisha app adds:
Free for ten days, then a subscription. We would rather write that here than have you find it later.